AGI will arrive within 2–4 years, and superintelligence will save Ben Goertzel’s 82-year-old parents from death, according to Goertzel’s own forecast

AGI will arrive within 2–4 years, and superintelligence will save Ben Goertzel’s 82-year-old parents from death, according to Goertzel’s own forecast

The founder of SingularityNET expects human-level AI within 2–4 years. The system will then be able to rewrite its own code and develop into a superintelligence within months or several years. He expects the consequences to include radical rejuvenation and freedom from the need to work.

Ben Goertzel helped bring the term “artificial general intelligence” into common use. On July 2, he gave an extended interview about the next few years. He defines general intelligence as the ability of an AI to learn different tasks and devise solutions beyond those already known.

He does not expect the next version of a conventional chatbot to reach this level. Goertzel is building a different architecture that combines a language model, logical reasoning, and a search for unexpected ideas, all guided by defined goals. Current AI can already write code and solve some mathematical problems better than humans. With its help, a development team can make as much progress in a year as it previously could in ten or twenty years.

Once a system reaches human-level intelligence, recursive self-improvement will begin. The AI will examine its own code, modify it, become more intelligent, and then use the improved version to find another way to enhance itself. Ray Kurzweil’s earlier forecast placed sixteen years between AGI in 2029 and the singularity. He used the term singularity for the point when the pace of discovery would become almost impossible for humans to comprehend. Goertzel expects the transition to take months or several years.

This expectation underlies his position on control. Preventing a single closed superintelligence from being copied would require total surveillance. Goertzel sees two realistic outcomes: AI controlled by corporations with ties to governments and militaries, or an open network similar to the internet. In SingularityNET and OpenBGI, different systems are intended to cooperate without a single owner.

Goertzel wants compassion and concern for sentient beings to be built into such a system’s initial goals. These values would then shape its learning and choice of actions from the beginning. Humans would still define the initial values. He therefore proposes creating many AI systems with different values, allowing them to cooperate and gradually revise those values over time.

The conversation then turns to his parents. They are both 82 years old and healthy. If superintelligence arrives within five years, “they probably won’t die”, Goertzel says. He imagines a daily pill that would, within several months, restore an 82-year-old person’s body to roughly that of a thirty-year-old. His reasoning is that cells can already be reprogrammed, so a superintelligence would only need to work out the biology of the human body.

His interviewer, Kai Bickle, says that immortality is not his goal. Goertzel asks, “At what age would you like to die?” Bickle replies that he is not afraid of death. “That isn’t what I asked,” Goertzel responds sharply. Bickle eventually chooses 120 healthy years, and Goertzel replies that he will change his mind again by then.

Goertzel also expects billions of people to be freed from repetitive work performed to obtain food and housing. In his view, this benefit outweighs the loss of professional purpose among an affluent minority. The transition will be painful for displaced specialists, but Goertzel expects AI to help provide healthcare, food, and infrastructure for most people.

Goertzel jokingly puts the probability that AI will cause human extinction at π: 3,14159%. He immediately describes this figure as intuition rather than calculation. In his view, the risk will be lower if AI learns compassion from the beginning instead of serving a military race between the United States and China.

Any path to such a pill will still require data, experiments, and clinical feedback, because the safety of a rejuvenation treatment must be tested in a living body. Goertzel explains his position simply: he wants to live a long time because he enjoys being alive. In his view, suicide at 120 years old would be unlikely to improve any possible afterlife compared with a life lasting ten million years.